FlashQLA v0.1.1

🚀 Highlights

• Intra‑card CP for the backward pass, developed by @Erix025
• SM100 architecture support: warp-specialized kernels with tcgen05
• Support state_v_first and aligned the entry function signature with the latest flash-linear-attention interface

🛠️ Additional Improvements

• Upgraded the tilelang dependency to v0.1.9
• Updated unit tests

📊 Benchmarks

• Benchmark results on SM90
• Benchmark results on SM100

🙏 Acknowledgements

Special thanks to all contributors and community members for their valuable feedback. We welcome your continued participation through issues and pull requests.